Unverified Commit a43b362e authored by Tobias Koppers's avatar Tobias Koppers Committed by GitHub

Turbopack: parallelize making dependent tasks dirty (#86755)

<!-- Thanks for opening a PR! Your contribution is much appreciated.
To make sure your PR is handled as smoothly as possible we request that you follow the checklist sections below.
Choose the right checklist for the change(s) that you're making:

## For Contributors

### Improving Documentation

- Run `pnpm prettier-fix` to fix formatting issues before opening the PR.
- Read the Docs Contribution Guide to ensure your contribution follows the docs guidelines: https://nextjs.org/docs/community/contribution-guide

### Fixing a bug

- Related issues linked using `fixes #number`
- Tests added. See: https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs
- Errors have a helpful link attached, see https://github.com/vercel/next.js/blob/canary/contributing.md

### Adding a feature

- Implements an existing feature request or RFC. Make sure the feature request has been accepted for implementation before opening a PR. (A discussion must be opened, see https://github.com/vercel/next.js/discussions/new?category=ideas)
- Related issues/discussions are linked using `fixes #number`
- e2e tests added (https://github.com/vercel/next.js/blob/canary/contributing/core/testing.md#writing-tests-for-nextjs)
- Documentation added
- Telemetry added. In case of a feature if it's used or not.
- Errors have a helpful link attached, see https://github.com/vercel/next.js/blob/canary/contributing.md


## For Maintainers

- Minimal description (aim for explaining to someone not on the team to understand the PR)
- When linking to a Slack thread, you might want to share details of the conclusion
- Link both the Linear (Fixes NEXT-xxx) and the GitHub issues
- Add review comments if necessary to explain to the reviewer the logic behind a change

### What?

### Why?

### How?

Closes NEXT-
Fixes #

-->
parent 6ed4453c
......@@ -34,10 +34,11 @@ use turbo_tasks::{
event::{Event, EventListener},
message_queue::TimingEvent,
registry::get_value_type,
scope::scope_and_block,
task_statistics::TaskStatisticsApi,
trace::TraceRawVcs,
turbo_tasks,
util::IdFactoryWithReuse,
util::{IdFactoryWithReuse, good_chunk_size, into_chunks},
};
pub use self::{operation::AnyOperation, storage::TaskDataCategory};
......@@ -46,10 +47,11 @@ use crate::backend::operation::TaskDirtyCause;
use crate::{
backend::{
operation::{
AggregationUpdateJob, AggregationUpdateQueue, CleanupOldEdgesOperation,
ComputeDirtyAndCleanUpdate, ConnectChildOperation, ExecuteContext, ExecuteContextImpl,
Operation, OutdatedEdge, TaskGuard, connect_children, get_aggregation_number,
get_uppers, is_root_node, make_task_dirty_internal, prepare_new_children,
AggregationUpdateJob, AggregationUpdateQueue, ChildExecuteContext,
CleanupOldEdgesOperation, ComputeDirtyAndCleanUpdate, ConnectChildOperation,
ExecuteContext, ExecuteContextImpl, Operation, OutdatedEdge, TaskGuard,
connect_children, get_aggregation_number, get_uppers, is_root_node,
make_task_dirty_internal, prepare_new_children,
},
storage::{
InnerStorageSnapshot, Storage, count, get, get_many, get_mut, get_mut_or_insert_with,
......@@ -68,6 +70,11 @@ use crate::{
},
};
/// Threshold for parallelizing making dependent tasks dirty.
/// If the number of dependent tasks exceeds this threshold,
/// the operation will be parallelized.
const DEPENDENT_TASKS_DIRTY_PARALLIZATION_THRESHOLD: usize = 10000;
const SNAPSHOT_REQUESTED_BIT: usize = 1 << (usize::BITS - 1);
/// Configurable idle timeout for snapshot persistence.
......@@ -2186,8 +2193,12 @@ impl<B: BackingStorage> TurboTasksBackendInner<B> {
);
}
let mut queue = AggregationUpdateQueue::new();
for dependent_task_id in output_dependent_tasks {
fn process_output_dependents(
ctx: &mut impl ExecuteContext<'_>,
task_id: TaskId,
dependent_task_id: TaskId,
queue: &mut AggregationUpdateQueue,
) {
#[cfg(feature = "trace_task_output_dependencies")]
let span = tracing::trace_span!(
"invalidate output dependency",
......@@ -2200,7 +2211,7 @@ impl<B: BackingStorage> TurboTasksBackendInner<B> {
// once tasks are never invalidated
#[cfg(feature = "trace_task_output_dependencies")]
span.record("result", "once task");
continue;
return;
}
let mut make_stale = true;
let dependent = ctx.task(dependent_task_id, TaskDataCategory::All);
......@@ -2217,7 +2228,7 @@ impl<B: BackingStorage> TurboTasksBackendInner<B> {
// output anymore and doesn't need to be invalidated
#[cfg(feature = "trace_task_output_dependencies")]
span.record("result", "no backward dependency");
continue;
return;
}
make_task_dirty_internal(
dependent,
......@@ -2225,14 +2236,41 @@ impl<B: BackingStorage> TurboTasksBackendInner<B> {
make_stale,
#[cfg(feature = "trace_task_dirty")]
TaskDirtyCause::OutputChange { task_id },
&mut queue,
queue,
ctx,
);
#[cfg(feature = "trace_task_output_dependencies")]
span.record("result", "marked dirty");
}
queue.execute(ctx);
if output_dependent_tasks.len() > DEPENDENT_TASKS_DIRTY_PARALLIZATION_THRESHOLD {
let chunk_size = good_chunk_size(output_dependent_tasks.len());
let chunks = into_chunks(output_dependent_tasks.to_vec(), chunk_size);
let _ = scope_and_block(chunks.len(), |scope| {
for chunk in chunks {
let child_ctx = ctx.child_context();
scope.spawn(move || {
let mut ctx = child_ctx.create();
let mut queue = AggregationUpdateQueue::new();
for dependent_task_id in chunk {
process_output_dependents(
&mut ctx,
task_id,
dependent_task_id,
&mut queue,
)
}
queue.execute(&mut ctx);
});
}
});
} else {
let mut queue = AggregationUpdateQueue::new();
for dependent_task_id in output_dependent_tasks {
process_output_dependents(ctx, task_id, dependent_task_id, &mut queue);
}
queue.execute(ctx);
}
}
fn task_execution_completed_unfinished_children_dirty(
......
......@@ -125,10 +125,10 @@ pub fn connect_children(
// This avoids long pauses of more than 30µs * 10k = 300ms.
// We don't want to parallelize too eagerly as spawning tasks and the temporary allocations have
// a cost as well.
const MIN_CHILDREN_FOR_PARALLEL: usize = 10000;
const CONNECT_CHILDREN_PARALLIZATION_THRESHOLD: usize = 10000;
let len = new_follower_ids.len();
if len >= MIN_CHILDREN_FOR_PARALLEL {
if len >= CONNECT_CHILDREN_PARALLIZATION_THRESHOLD {
let new_follower_ids = new_follower_ids.into_vec();
let chunk_size = good_chunk_size(len);
let _ = scope_and_block(len.div_ceil(chunk_size), |scope| {
......
......@@ -302,6 +302,16 @@ pub struct IntoChunks<T> {
chunk_size: usize,
}
impl<T> IntoChunks<T> {
pub fn len(&self) -> usize {
(self.data.len() - self.index).div_ceil(self.chunk_size)
}
pub fn is_empty(&self) -> bool {
self.index >= self.data.len()
}
}
impl<T> Iterator for IntoChunks<T> {
type Item = Chunk<T>;
......@@ -386,6 +396,18 @@ impl<T> Drop for Chunk<T> {
mod tests {
use super::*;
#[test]
fn test_into_chunks_len() {
assert_eq!(into_chunks::<i32>(vec![], 2).len(), 0);
assert_eq!(into_chunks(vec![1], 2).len(), 1);
assert_eq!(into_chunks(vec![1, 2], 2).len(), 1);
assert_eq!(into_chunks(vec![1, 2, 3], 2).len(), 2);
assert_eq!(into_chunks(vec![1, 2, 3, 4], 2).len(), 2);
assert_eq!(into_chunks(vec![1, 2, 3, 4, 5], 2).len(), 3);
assert_eq!(into_chunks(vec![1, 2, 3, 4, 5, 6], 2).len(), 3);
assert_eq!(into_chunks(vec![1, 2, 3, 4, 5, 6, 7], 2).len(), 4);
}
#[test]
fn test_chunk_iterator() {
let data = [(); 10]
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment